[HDFS-8339] Erasure Coding: Badly treated when createBlockOutputStream failed in DataStreamer

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Duplicate
• Assignee: Unassigned
• Reporter: Walter Su
• Created: 07/May/15 11:36
• Updated: 18/May/15 09:21
• Resolved: 18/May/15 09:21

Description

Issue 1:

leading streamer calls abandonBlock(..) and get new blockGroup from NN, primary streamer should sync with non-leading streamer instead of throw exception( offer twice to stripedBlock blockingQueue).

2015-05-07 18:58:05,335 INFO  hdfs.DataStreamer (DataStreamer.java:nextBlockOutputStream(1386)) - Abandoning BP-172584615-9.96.1.34-1430996280714:blk_-9223372036854775792_1001
...
2015-05-07 18:58:05,373 WARN  hdfs.DataStreamer (DataStreamer.java:run(572)) - DataStreamer Exception
java.io.IOException: Failed: LocatedBlock{BP-172584615-9.96.1.34-1430996280714:blk_-9223372036854775770_1002; getBlockSize()=0; corrupt=false; offset=1572864; locs=[DatanodeInfoWithStorage[127.0.0.1:52490,DS-6080b76f-adf7-45a8-aa0e-e0e82c2c1569,DISK]]}, i=6
    at org.apache.hadoop.hdfs.DFSStripedOutputStream$Coordinator.putStripedBlock(DFSStripedOutputStream.java:117)
    at org.apache.hadoop.hdfs.StripedDataStreamer.locateFollowingBlock(StripedDataStreamer.java:120)
    at org.apache.hadoop.hdfs.DataStreamer.nextBlockOutputStream(DataStreamer.java:1364)
    at org.apache.hadoop.hdfs.DataStreamer.run(DataStreamer.java:461)
    at org.apache.hadoop.hdfs.StripedDataStreamer.run(StripedDataStreamer.java:48)

Issue 2:

non-leading streamer calls abandonBlock(..) and get new locatedBlock from coordinator. Actually it's the last blockGroup, no more locatedBlock can poll from stripedBlocks blockingQueue. Other 8 streamer finished and closed, but this streamer hangs about 90 seconds.

2015-05-07 19:21:25,357 INFO  BlockStateChange (BlockManager.java:logAddStoredBlock(2768)) - BLOCK* addStoredBlock: blockMap updated: 127.0.0.1:51998 is added to ...
2015-05-07 19:22:55,250 WARN  hdfs.DataStreamer (DataStreamer.java:run(572)) - DataStreamer Exception
java.io.IOException: Failed: i=1 
    at org.apache.hadoop.hdfs.DFSStripedOutputStream$Coordinator.getStripedBlock(DFSStripedOutputStream.java:130)
    at org.apache.hadoop.hdfs.StripedDataStreamer.locateFollowingBlock(StripedDataStreamer.java:124)
    at org.apache.hadoop.hdfs.DataStreamer.nextBlockOutputStream(DataStreamer.java:1364)
    at org.apache.hadoop.hdfs.DataStreamer.run(DataStreamer.java:461)
    at org.apache.hadoop.hdfs.StripedDataStreamer.run(StripedDataStreamer.java:48)

Issue 3:

remove abandonBlock(..) RPC call for non-leading streamer

Issue Links

• is duplicated by: HDFS-8383 Tolerate multiple failures in DFSStripedOutputStream
• relates to: HDFS-7672 Erasure Coding: consolidate streamer coordination logic and handle failure when writing striped blocks